On the dark side
To optimize the stroke-written character set for the application, we initially selected a 2716 UV (ultraviolet) EPROM, which would enable us to experiment and quickly make changes. One of the engineers worked on the character set and programmed a 2716 for the characters we needed. We tested it, and everything worked fine. We soon realized, however, that the recently introduced 2732 EPROM would probably obsolete the 2716. It was still early in the design cycle, so we decided to use the 2732 instead. It came in the same package and had the same pinout as the 2716 except for an extra address bit, so it was an easy change. We programmed a 2732 and inserted it in the socket on the breadboard. Everything worked fine. We continued testing other aspects of the design, working out other bugs here and there.
It was getting closer to a major design review. Our customer would be visiting and expected to see the prototype in operation. We were confident that the demo would go flawlessly because we had tested the heck out of it. What could go wrong? For the demonstration, we wanted to show the customer that an operator could dim the display to low light for nighttime operation. Although we had measured the light output and confirmed that it was in spec, the numbers on the test instrumentation were less impressive than a demo would show. We asked facilities to build us a “darkroom” in the lab so that we could demonstrate the display in both bright-sunlight and dark conditions.
After some thought, we realized that the original schematic wired up only the address lines that the 2716 used and that we had forgotten to wire the extra address line of the 2732. Yet the 2732 had been running fine in the lab for months. We then realized that the lab always had light coming through the quartz window that allowed the UV EPROM to be erased. We had never bothered to cover the window because we were still in development mode, and the light in the lab was too dim to erase the UV EPROM in any reasonable time. The extra address line on the 2732 happened to go to a low logic level in the light but drifted to a high logic level in the dark. That action would select the half of the EPROM that we had not programmed, which also happened to be the program for all space characters. Therefore, the entire display went blank.
Tales from the Cube
The next day, one of the engineers called me into the darkroom and informed me of a new problem. When he turned off the light, the entire display changed to Chinese characters. When he flipped the switch back on, everything returned to English. I realized that, in a celebratory mood after our successful demo, he was pulling a prank on me. He had programmed half of the 2732 with Chinese characters and wired up the floating address line with a switch to either let the address line float or connect it to ground. We both laughed and went back to work.
Walter Sjursen is the chief technology officer at Songbird Hearing Inc (North Brunswick, NJ).